智能的本质与意义OA
On intelligence:the origin and the goal
以LLM把通识问题的问答变成随口可得为起点,指出智业时代已然到来,教育、研究与工作方式将被整体改写,人类也因此面临"无田""无产""无用"的焦虑与危机.为回答这一时代挑战,从复杂性与智能化的对立统一谈起:生命可视为复杂化的物质,智能可视为复杂化的计算,复杂性的多样性与不定性决定了智能涌现与治理的根本难题.进而借助对立统一的思想线索,提出向虚而生的平行路径,并以ACP方法把对立置于人工社会、把统一落实于实际社会,通过计算实验与平行执行实现从小知识到大模型,再到深智能的闭环生成.最后以3个世界的本体框架重述老旧新3类AI的演化趋势,强调必须在发展中同步推进安全与治理,使智业时代依然做到人智为基、万智之擎,造福而非危害人类.
Large language models now provide instant answers to general-education questions.Starting from this observa-tion,this article argues that the intelligent era has arrived.This era reshapes education,research,and work.Consequently,people face anxieties and crises of transitioning from being landless and propertyless to feeling useless.To address these challenges,the discussion starts with the unity of opposites between complexity and intelligence.Life can be viewed as complexified matter,and intelligence as complexified computation.The diversities and uncertainties of complexity deter-mine the fundamental difficulties in the emergence and governance of intelligence.Following this line of thought,the ar-ticle proposes a parallel path of generating substance from cyberspace.The ACP approach places oppositions within artifi-cial societies and implements unities within actual societies.Through computational experiments and parallel executions,this framework forms closed loops that transform small knowledge into large models,and subsequently into deep intelli-gence.Finally,utilizing the ontological framework of three worlds,the article recounts the evolutionary trends of"old","past",and"new"AI.It emphasizes that safety and governance must advance synchronously with development.This en-sures the intelligent era remains founded on human value and powered by diverse intelligences,ultimately benefiting hu-manity.
王飞跃
中国科学院自动化研究所复杂系统管理与控制国家重点实验室,北京 100190||澳门科技大学创新工程学院工程科学系,澳门 999078
信息技术与安全科学
智能本质智能ACP人工智能平行智能智能科学智能技术知识自动化
intelligenceintelligence complexityACPartificial intelligenceparallel intelligenceintelligent sciencein-telligent technologyknowledge automation
《智能科学与技术学报》 2026 (1)
1-11,11
评论